Anthropic’s first public filing shows the financial pressures behind its rapid growth, its reliance on big clients and the risks its standard artificial intelligence (AI) technology may pose.
Artificial intelligence company Anthropic has filed initial papers for a planned US stock market debut, giving investors a detailed look at the business behind its Claude chatbot.
A 335-page registration statement filed with the US Securities and Exchange Commission on October 26, 2026 details the company's financial performance, commitments to infrastructure, customer relationships and assessment of the risks of advanced AI. This filing is preliminary, and the terms of the proposed offering have not been determined.
The disclosure raises five key issues.
1. Advanced AI could be an existential risk, says Anthropic.
Anthropic’s risk disclosures go beyond those typical of a business to the possibility that advanced artificial intelligence systems could pose a much broader threat to human safety.
The company warns that systems sufficiently advanced could present catastrophic or existential risks if they are developed or deployed without appropriate safeguards, oversight and security measures.
Its filing also suggests the possibility of AI being used in cyberattacks, the risk that technological capabilities could outrun safety precautions, and the threat of serious incidents undermining public trust in the industry.
Anthropic’s business also relies on its ability to build ever more capable models, while still having access to the computing infrastructure necessary to train and run them.
This infrastructure includes specialised chips, data centres and cloud computing services. Disruption to these resources could impact the company’s ability to meet customer demand and compete with other AI developers.
The disclosures show the conundrum the industry faces: companies must expand the technology’s capabilities while minimising risks that could be harder to contain as the capabilities expand.
2. Further scrutiny is needed of the $80.7bn headline loss.
For the six months to June 30, 2026, Anthropic posted a net loss of about $80.7bn on revenue of about $16.3bn.
The loss is stark, but it is not $80.7bn spent on day-to-day operations.
Some $72.4 billion of the loss related to changes in the fair value of convertible notes, derivative liabilities and other financing instruments, the company's filing said. The non-cash accounting charges were primarily related to increases in the estimated fair value of the company and its outstanding convertible financing arrangements.
That distinction is important in evaluating the company’s financial position.
Anthropic’s operating performance paints a different picture to its headline net loss. The company reported an operating loss of $693m in the first half of 2026, while adjusted operating income was positive at $1.62bn, Reuters reported. These are based on different accounting adjustments and should not be considered interchangeable. Turn 534697 News 25
Revenue growth has been rapid, but profitability is a key consideration for potential investors. The company has to continue funding research, staffing and infrastructure while managing the financial obligations of its expansion.
The proposed listing is expected to attract significant interest from investors, although its ultimate valuation will depend on market conditions, the final terms of the offering and demand from investors.
3. Massive infrastructure commitments for huge cash reserves
Anthropic had about $131.7bn in cash and short-term investments as of June 30, 2026. That gives a huge amount of financial resources to play with but must be seen in conjunction with the company's future spending commitments.
The company is putting money into the computing power required to build and run advanced AI models. It offers cloud services, specialised hardware and data centre infrastructure.
The filing showed estimated future lease obligations of approximately $321.4bn for leases that had not commenced as of June 30, 2026. It also reported material purchase commitments for computing and infrastructure resources. These are different types of contractual commitments and should not be added together without checking for overlaps.
While such commitments may help to ensure access to computing resources as demand grows, they also create long-term financial obligations.
So Anthropic needs to balance having enough infrastructure for its business but not having fixed costs grow faster than sustainable revenue growth.
Its cash reserves give it flexibility but do not mitigate the risks of long-term commitments, changing technology needs or increased competition.
4. A small number of customers account for a large share of the business.
Anthropic has built momentum in its enterprise business, with organisations embracing AI tools for software development, research, customer support and other tasks.
But this dependence on a smallish number of big customers also leaves the company potentially exposed.
Source material figures suggest that 250 enterprise customers each paid at least $10m in the last year, and five customers accounted for around a quarter of the business. These should be cross-checked with the relevant definitions and period in the final prospectus before publication.
Revenue could be vulnerable to customer concentration if a major client cuts back spending, switches providers or decides to build more of its own AI capabilities.
Then there is the question of contract terms. Future revenue could be less predictable because major customers do not have long-term agreements. Some of these customers have not entered into long-term agreements.
Anthropic also has commercial relationships with large technology companies that invest in the business and compete in overlapping markets. These relationships can offer distribution and infrastructure advantages but can also lead to competing commercial interests.
The company’s strategy is to get Claude in users’ hands first, then try to convert them to paying customers. The approach could expand its customer base, but the financial return will depend on how many users convert to paying customers and how much revenue they bring in.
Anthropic argues that AI could provide significant public benefits.
Anthropic couples its warnings about AI risks with a broader argument for the technology’s potential benefits.
The company says advanced AI could accelerate scientific research, help improve disease detection, help develop treatments and help researchers make headway on complex problems.
It could also be used in education, financial and legal services, energy infrastructure and battery design.
These are possibilities, not certainties. The impact of AI in the real world depends on the reliability of artificial intelligence systems, the quality of data used, the costs of deployment and the safeguards governing their use.
Anthropic says its mission is to build advanced AI responsibly for humanity’s long-term benefit. Its filing shows that objective along with the commercial imperatives of building and operating ever more capable systems.
The big question for investors is whether the company can turn growing interest in AI into lasting profits, meet its infrastructure commitments and manage the risks of the technology.
The planned listing will offer public-market investors a glimpse of that challenge, but the preliminary filing does not set the expectation that Anthropic will meet its long-term financial or technology ambitions.
